DESCRIPTION: This proposal describes a plan to further study the determinants and consequences of substance use using data from a longitudinal study of urban adolescents. Previous studies investigating substance use and its associated comorbidities often focus on individual-level risk and protective factors. More recent work has envisioned individuals as part of a socio-ecological context, which motivates studying the effects of neighborhood- and community-level risk and protective factors and we intend to build on this on two ways. First, The transition from adolescence to adulthood is a period of rapid biological, cognitive and social development, presenting the possibility that the way an individual reacts to his or her environment, and therefore the effect it has, is age-dependent. Specifically, due to age-specific developmental and societal context, it is likely that rates of substance use, what variables operate as risk/protective factors for substance use, and the level of comorbidity between substance use and other negative behaviors are all age-dependent. Second, the same socio-ecological context that been found to modulate substance use rates may also affect the link between substance use and other negative behaviors. These dynamics have gone unexplored to this point and their exploration is the purpose of this proposal. To address this first gap, we propose using varying coefficient regression models to explore how the effects of individual- and community- level risk factors for substance use change with age. This will tell us not only what risk/protective factors for substance use are important, but when they are most important. To address the second gap, we will derive a previously unexplored statistical model of the dependence between two binary variables (e.g. substance use and weapon violence). In this model - whose implementation will be made freely available as an R package - we will use a combination of Bayesian and composite likelihood approaches to isolate estimation of the odds ratio - and how it changes as a function of predictor variables - without the need to estimate parameters of secondary interest, such as means or temporal autocorrelation parameters. Within this framework, we will determine the factors that increase/decrease the level of comorbidity between substance use and related negative behaviors, as well as how these relationships change with age.